kubeflow
/
examples
mirror of https://github.com/kubeflow/examples.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

update kfp 'github issue summarization' example (#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
This commit is contained in:
Amy 2020-10-06 05:13:43 -07:00 committed by GitHub
parent 31a4d5ea8a
commit 44d722222f
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
17 changed files with 251 additions and 198 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
github_issue_summarization/pipelines/README.md
View File

@ -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). -->

26
github_issue_summarization/pipelines/components/kubeflow-resources/containers/tf-serving-gh/Dockerfile
View File

@ -12,7 +12,7 @@
# 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
@ -20,7 +20,20 @@ RUN apt-get install --no-install-recommends -y -q ca-certificates python-dev pyt
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"]

5
github_issue_summarization/pipelines/components/kubeflow-resources/tf-serving-gh/deploy-tf-serve.py
View File

@ -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)

31
github_issue_summarization/pipelines/components/kubeflow-resources/tf-serving-gh/tf-serve-template.yaml
View File

@ -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

4
github_issue_summarization/pipelines/components/t2t/containers/t2t_app/Dockerfile
View File

@ -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

23
github_issue_summarization/pipelines/components/t2t/containers/webapp-launcher/Dockerfile
View File

@ -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

4
github_issue_summarization/pipelines/components/t2t/t2t-app/app/templates/index.html
View File

@ -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>

7
github_issue_summarization/pipelines/components/t2t/webapp-launcher/deploy-webapp.py
View File

@ -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)

55
github_issue_summarization/pipelines/components/t2t/webapp-launcher/t2tapp-template.yaml
View File

@ -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

BIN
github_issue_summarization/pipelines/example_pipelines/gh_summ.py.tar.gz
View File

Binary file not shown.

88
github_issue_summarization/pipelines/example_pipelines/gh_summ_hosted_kfp.py Normal file
View File

@ -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')

BIN
github_issue_summarization/pipelines/example_pipelines/gh_summ_hosted_kfp.py.tar.gz Normal file
View File

Binary file not shown.

66
github_issue_summarization/pipelines/example_pipelines/gh_summ_preempt.py → github_issue_summarization/pipelines/example_pipelines/gh_summ_hosted_kfp_preempt.py
View File

@ -16,6 +16,7 @@
import kfp.dsl as dsl
import kfp.gcp as gcp
import kfp.components as comp
# from kfp.dsl.types import GCSPath, String
COPY_ACTION = 'copy_data'
@ -32,77 +33,54 @@ 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
)
metadata_log_op = comp.load_component_from_url(
'https://raw.githubusercontent.com/kubeflow/examples/master/github_issue_summarization/pipelines/components/t2t/metadata_log_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=2019300,
project='YOUR_PROJECT_HERE',
github_token='YOUR_GITHUB_TOKEN_HERE',
working_dir='YOUR_GCS_DIR_HERE',
checkpoint_dir='gs://aju-dev-demos-codelabs/kubecon/model_output_tbase.bak2019000',
deploy_webapp='true',
data_dir='gs://aju-dev-demos-codelabs/kubecon/t2t_data_gh_all/'
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(
working_dir=working_dir,
data_dir=data_dir,
checkpoint_dir=checkpoint_dir,
model_dir='%s/%s/model_output' % (working_dir, '{{workflow.name}}'),
action=COPY_ACTION
).apply(gcp.use_gcp_secret('user-gcp-sa'))
log_dataset = metadata_log_op(
log_type=DATASET,
workspace_name=WORKSPACE_NAME,
run_name='{{workflow.name}}',
data_uri=data_dir
model_dir='%s/%s/model_output' % (working_dir, dsl.RUN_ID_PLACEHOLDER),
action=COPY_ACTION,
)
train = train_op(
working_dir=working_dir,
data_dir=data_dir,
checkpoint_dir=checkpoint_dir,
model_dir='%s/%s/model_output' % (working_dir, '{{workflow.name}}'),
model_dir=copydata.outputs['copy_output_path'],
action=TRAIN_ACTION, train_steps=train_steps,
deploy_webapp=deploy_webapp
).apply(gcp.use_gcp_secret('user-gcp-sa'))
log_model = metadata_log_op(
log_type=MODEL,
workspace_name=WORKSPACE_NAME,
run_name='{{workflow.name}}',
model_uri='%s/%s/model_output' % (working_dir, '{{workflow.name}}')
)
serve = dsl.ContainerOp(
name='serve',
image='gcr.io/google-samples/ml-pipeline-kubeflow-tfserve',
arguments=["--model_name", 'ghsumm-%s' % ('{{workflow.name}}',),
"--model_path", '%s/%s/model_output/export' % (working_dir, '{{workflow.name}}')
image='gcr.io/google-samples/ml-pipeline-kubeflow-tfserve:v5',
arguments=["--model_name", 'ghsumm-%s' % (dsl.RUN_ID_PLACEHOLDER,),
"--model_path", train.outputs['train_output_path']
]
)
log_dataset.after(copydata)
train.after(copydata)
log_model.after(train)
serve.after(train)
train.set_gpu_limit(4).apply(gcp.use_preemptible_nodepool()).set_retry(5)
train.set_memory_limit('48G')
with dsl.Condition(train.output == 'true'):
# train.set_gpu_limit(1)
train.set_gpu_limit(1).apply(gcp.use_preemptible_nodepool()).set_retry(10)
with dsl.Condition(train.outputs['launch_server'] == 'true'):
webapp = dsl.ContainerOp(
name='webapp',
image='gcr.io/google-samples/ml-pipeline-webapp-launcher:v2ap',
arguments=["--model_name", 'ghsumm-%s' % ('{{workflow.name}}',),
image='gcr.io/google-samples/ml-pipeline-webapp-launcher:v7ap',
arguments=["--model_name", 'ghsumm-%s' % (dsl.RUN_ID_PLACEHOLDER,),
"--github_token", github_token]
)

BIN
github_issue_summarization/pipelines/example_pipelines/gh_summ_serve.py.tar.gz
View File

Binary file not shown.

0
github_issue_summarization/pipelines/example_pipelines/gh_summ.py → github_issue_summarization/pipelines/example_pipelines/obsolete/gh_summ.py
View File

0
github_issue_summarization/pipelines/example_pipelines/gh_summ_serve.py → github_issue_summarization/pipelines/example_pipelines/obsolete/gh_summ_serve.py
View File

138
github_issue_summarization/pipelines/example_pipelines/pipelines-notebook.ipynb
View File

@ -6,7 +6,7 @@
"source": [
"# KubeFlow Pipeline: Github Issue Summarization using Tensor2Tensor\n",
"\n",
"This notebook assumes that you have already set up a GKE cluster with Kubeflow installed as per this codelab: [g.co/codelabs/kubecon18](g.co/codelabs/kubecon18). Currently, this notebook must be run from the Kubeflow JupyterHub installation, as described in the codelab.\n",
"This notebook assumes that you have already set up a GKE cluster with CAIP Pipelines (Hosted KFP) installed, with the addition of a GPU-enabled node pool, as per this codelab: [g.co/codelabs/kubecon18](g.co/codelabs/kubecon18).\n",
"\n",
"In this notebook, we will show how to:\n",
"\n",
@ -29,7 +29,7 @@
"cell_type": "markdown",
"metadata": {},
"source": [
"Do some installations and imports, and set some variables. Set the `WORKING_DIR` to a path under the Cloud Storage bucket you created earlier. The Pipelines SDK is bundled with the notebook server image, but we'll make sure that we're using the most current version for this example. You may need to restart your kernel after the SDK update."
"Do some installations and imports, and set some variables. Set the `WORKING_DIR` to a path under the Cloud Storage bucket you created earlier. You may need to restart your kernel after the KFP SDK update."
]
},
{
@ -41,6 +41,18 @@
"!pip install -U kfp"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# Restart kernel after the pip install\n",
"import IPython\n",
"\n",
"IPython.Application.instance().kernel.do_shutdown(True)"
]
},
{
"cell_type": "code",
"execution_count": null,
@ -65,19 +77,37 @@
"source": [
"# Define some pipeline input variables. \n",
"WORKING_DIR = 'gs://YOUR_GCS_BUCKET/t2t/notebooks' # Such as gs://bucket/object/path\n",
"PROJECT_NAME = 'YOUR_PROJECT'\n",
"GITHUB_TOKEN = 'YOUR_GITHUB_TOKEN' # needed for prediction, to grab issue data from GH\n",
"\n",
"DEPLOY_WEBAPP = 'false'"
"PROJECT_NAME = 'YOUR_PROJECT'\n",
"GITHUB_TOKEN = 'YOUR_GITHUB_TOKEN' # optional; used for prediction, to grab issue data from GH\n",
"\n",
"DEPLOY_WEBAPP = 'false' # change this to 'true' to deploy a new version of the webapp part of the pipeline"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Create an *Experiment* in the Kubeflow Pipeline System\n",
"## Instantiate the KFP client and create an *Experiment* in the Kubeflow Pipeline System\n",
"\n",
"The Kubeflow Pipeline system requires an \"Experiment\" to group pipeline runs. You can create a new experiment, or call `client.list_experiments()` to get existing ones."
"Next we'll instantiate a KFP client object with the `host` info from your Hosted KFP installation. To do this, go to the Pipelines dashboard in the Cloud Console and click on the \"Settings\" gear for the KFP installation that you want to use. You'll see a popup window. Look for the \"Connect to this Kubeflow Pipelines instance...\" text and copy the \"client = kfp.Client(...)\" line below it. Edit the following cell to use that line."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# CHANGE THIS with the info for your KFP cluster installation\n",
"client = kfp.Client(host='xxxxxxxx-dot-us-centralx.pipelines.googleusercontent.com')"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The Kubeflow Pipeline system requires an \"Experiment\" to group pipeline runs. You can create a new experiment, or call `client.list_experiments()` to get existing ones. (This will also serve to check that your client is set up properly)."
]
},
{
@ -86,9 +116,6 @@
"metadata": {},
"outputs": [],
"source": [
"# Note that this notebook should be running in JupyterHub in the same cluster as the pipeline system.\n",
"# Otherwise, additional config would be required to connect.\n",
"client = kfp.Client()\n",
"client.list_experiments()"
]
},
@ -136,7 +163,6 @@
"source": [
"COPY_ACTION = 'copy_data'\n",
"TRAIN_ACTION = 'train'\n",
"WORKSPACE_NAME = 'ws_gh_summ'\n",
"DATASET = 'dataset'\n",
"MODEL = 'model'\n",
"\n",
@ -146,11 +172,7 @@
"\n",
"train_op = comp.load_component_from_url(\n",
" 'https://raw.githubusercontent.com/kubeflow/examples/master/github_issue_summarization/pipelines/components/t2t/train_component.yaml' # pylint: disable=line-too-long\n",
" )\n",
"\n",
"metadata_log_op = comp.load_component_from_url(\n",
" 'https://raw.githubusercontent.com/kubeflow/examples/master/github_issue_summarization/pipelines/components/t2t/metadata_log_component.yaml' # pylint: disable=line-too-long\n",
" )"
" )\n"
]
},
{
@ -166,35 +188,25 @@
"metadata": {},
"outputs": [],
"source": [
"\n",
"@dsl.pipeline(\n",
" name='Github issue summarization',\n",
" description='Demonstrate Tensor2Tensor-based training and TF-Serving'\n",
")\n",
"def gh_summ( #pylint: disable=unused-argument\n",
"def gh_summ(\n",
" train_steps: 'Integer' = 2019300,\n",
" project: str = 'YOUR_PROJECT_HERE',\n",
" github_token: str = 'YOUR_GITHUB_TOKEN_HERE',\n",
" working_dir: 'GCSPath' = 'YOUR_GCS_DIR_HERE',\n",
" working_dir: 'GCSPath' = 'gs://YOUR_GCS_DIR_HERE',\n",
" checkpoint_dir: 'GCSPath' = 'gs://aju-dev-demos-codelabs/kubecon/model_output_tbase.bak2019000/',\n",
" deploy_webapp: str = 'true',\n",
" data_dir: 'GCSPath' = 'gs://aju-dev-demos-codelabs/kubecon/t2t_data_gh_all/'\n",
" ):\n",
"\n",
"\n",
" copydata = copydata_op(\n",
" data_dir=data_dir,\n",
" checkpoint_dir=checkpoint_dir,\n",
" model_dir='%s/%s/model_output' % (working_dir, dsl.RUN_ID_PLACEHOLDER),\n",
" action=COPY_ACTION,\n",
" ).apply(gcp.use_gcp_secret('user-gcp-sa'))\n",
"\n",
"\n",
" log_dataset = metadata_log_op(\n",
" log_type=DATASET,\n",
" workspace_name=WORKSPACE_NAME,\n",
" run_name=dsl.RUN_ID_PLACEHOLDER,\n",
" data_uri=data_dir\n",
" )\n",
"\n",
" train = train_op(\n",
@ -202,32 +214,22 @@
" model_dir=copydata.outputs['copy_output_path'],\n",
" action=TRAIN_ACTION, train_steps=train_steps,\n",
" deploy_webapp=deploy_webapp\n",
" ).apply(gcp.use_gcp_secret('user-gcp-sa'))\n",
"\n",
"\n",
" log_model = metadata_log_op(\n",
" log_type=MODEL,\n",
" workspace_name=WORKSPACE_NAME,\n",
" run_name=dsl.RUN_ID_PLACEHOLDER,\n",
" model_uri=train.outputs['train_output_path']\n",
" )\n",
"\n",
" serve = dsl.ContainerOp(\n",
" name='serve',\n",
" image='gcr.io/google-samples/ml-pipeline-kubeflow-tfserve',\n",
" image='gcr.io/google-samples/ml-pipeline-kubeflow-tfserve:v5',\n",
" arguments=[\"--model_name\", 'ghsumm-%s' % (dsl.RUN_ID_PLACEHOLDER,),\n",
" \"--model_path\", train.outputs['train_output_path']\n",
" ]\n",
" )\n",
" log_dataset.after(copydata)\n",
" log_model.after(train)\n",
"\n",
" train.set_gpu_limit(1)\n",
" train.set_memory_limit('48G')\n",
"\n",
" with dsl.Condition(train.outputs['launch_server'] == 'true'):\n",
" webapp = dsl.ContainerOp(\n",
" name='webapp',\n",
" image='gcr.io/google-samples/ml-pipeline-webapp-launcher:v2ap',\n",
" image='gcr.io/google-samples/ml-pipeline-webapp-launcher:v7ap',\n",
" arguments=[\"--model_name\", 'ghsumm-%s' % (dsl.RUN_ID_PLACEHOLDER,),\n",
" \"--github_token\", github_token]\n",
"\n",
@ -302,7 +304,7 @@
" name='datagen',\n",
" image='gcr.io/google-samples/ml-pipeline-t2tproc',\n",
" arguments=[ \"--data-dir\", data_dir, \"--project\", project]\n",
" ).apply(gcp.use_gcp_secret('user-gcp-sa'))"
" )"
]
},
{
@ -346,14 +348,6 @@
" checkpoint_dir=checkpoint_dir,\n",
" model_dir='%s/%s/model_output' % (working_dir, dsl.RUN_ID_PLACEHOLDER),\n",
" action=COPY_ACTION,\n",
" ).apply(gcp.use_gcp_secret('user-gcp-sa'))\n",
"\n",
"\n",
" log_dataset = metadata_log_op(\n",
" log_type=DATASET,\n",
" workspace_name=WORKSPACE_NAME,\n",
" run_name=dsl.RUN_ID_PLACEHOLDER,\n",
" data_uri=data_dir\n",
" )\n",
"\n",
" train = train_op(\n",
@ -361,38 +355,28 @@
" model_dir=copydata.outputs['copy_output_path'],\n",
" action=TRAIN_ACTION, train_steps=train_steps,\n",
" deploy_webapp=deploy_webapp\n",
" ).apply(gcp.use_gcp_secret('user-gcp-sa'))\n",
"\n",
" log_dataset.after(copydata)\n",
" train.after(preproc)\n",
"\n",
" log_model = metadata_log_op(\n",
" log_type=MODEL,\n",
" workspace_name=WORKSPACE_NAME,\n",
" run_name=dsl.RUN_ID_PLACEHOLDER,\n",
" model_uri=train.outputs['train_output_path']\n",
" )\n",
" train.after(preproc) \n",
"\n",
" serve = dsl.ContainerOp(\n",
" name='serve',\n",
" image='gcr.io/google-samples/ml-pipeline-kubeflow-tfserve',\n",
" image='gcr.io/google-samples/ml-pipeline-kubeflow-tfserve:v5',\n",
" arguments=[\"--model_name\", 'ghsumm-%s' % (dsl.RUN_ID_PLACEHOLDER,),\n",
" \"--model_path\", train.outputs['train_output_path']\n",
" ]\n",
" )\n",
" log_model.after(train)\n",
"\n",
" train.set_gpu_limit(1)\n",
" train.set_memory_limit('48G')\n",
"\n",
" with dsl.Condition(train.outputs['launch_server'] == 'true'):\n",
" webapp = dsl.ContainerOp(\n",
" name='webapp',\n",
" image='gcr.io/google-samples/ml-pipeline-webapp-launcher:v2ap',\n",
" image='gcr.io/google-samples/ml-pipeline-webapp-launcher:v7ap',\n",
" arguments=[\"--model_name\", 'ghsumm-%s' % (dsl.RUN_ID_PLACEHOLDER,),\n",
" \"--github_token\", github_token]\n",
"\n",
" )\n",
" webapp.after(serve)\n"
" webapp.after(serve)"
]
},
{
@ -429,13 +413,20 @@
"metadata": {},
"source": [
"You should be able to see your newly defined pipeline run in the dashboard:\n",
"![](https://storage.googleapis.com/amy-jo/images/kf-pls/Screenshot%202019-09-04%2013.24.19%202.png)\n",
"\n",
"![](https://storage.googleapis.com/amy-jo/images/kf-pls/t2t_pipeline_in_dashboard.png)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The new pipeline has the following structure:\n",
"![The new pipeline structure.](https://storage.googleapis.com/amy-jo/images/kf-pls/pipeline_new.png)\n",
"\n",
"![The new pipeline structure.](https://storage.googleapis.com/amy-jo/images/kf-pls/t2t_pipeline_structure.png)\n",
"\n",
"Below is a screenshot of the pipeline running.\n",
"![The pipeline running.](https://storage.googleapis.com/amy-jo/images/kf-pls/Screenshot_2019-09-06_13_06_50.png)"
"\n",
"![The pipeline running.](https://storage.googleapis.com/amy-jo/images/kf-pls/t2t_pipeline_running.png)"
]
},
{
@ -467,6 +458,11 @@
}
],
"metadata": {
"environment": {
"name": "tf2-2-2-gpu.2-2.m48",
"type": "gcloud",
"uri": "gcr.io/deeplearning-platform-release/tf2-2-2-gpu.2-2:m48"
},
"kernelspec": {
"display_name": "Python 3",
"language": "python",
@ -482,9 +478,9 @@
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.6.7"
"version": "3.7.6"
}
},
"nbformat": 4,
"nbformat_minor": 2
"nbformat_minor": 4
}