kserve/docs/samples/pipelines/kfs-pipeline.ipynb

KServe Pipeline samples

This notebook assumes your cluster has KServe >= v0.7.0 installed which supports the v1beta1 API.

Install the necessary kfp library

In [ ]:

!pip3 install kfp --upgrade

In [ ]:

import kfp.compiler as compiler
import kfp.dsl as dsl
import kfp
from kfp import components

# Create kfp client
# Note: Add the KubeFlow Pipeline endpoint below if the client is not running on the same cluster.
# Example: kfp.Client('http://192.168.1.27:31380/pipeline')
client = kfp.Client()
EXPERIMENT_NAME = "KServe Experiments"
experiment = client.create_experiment(name=EXPERIMENT_NAME, namespace="anonymous")

TensorFlow example

In [ ]:

# kfserving_op = components.load_component_from_url('https://raw.githubusercontent.com/kubeflow/pipelines/master/components/kubeflow/kfserving/component.yaml')
kserve_op = components.load_component_from_url(
    "https://raw.githubusercontent.com/kubeflow/pipelines/master/components/kserve/component.yaml"
)


@dsl.pipeline(name="KServe pipeline", description="A pipeline for KServe.")
def kservePipeline(
    action="apply",
    model_name="tensorflow-sample",
    model_uri="gs://kfserving-examples/models/tensorflow/flowers",
    namespace="anonymous",
    framework="tensorflow",
):

    kserve = kserve_op(
        action=action,
        model_name=model_name,
        model_uri=model_uri,
        namespace=namespace,
        framework=framework,
    ).set_image_pull_policy("Always")


# Compile pipeline
compiler.Compiler().compile(kservePipeline, "tf-flower.tar.gz")

# Execute pipeline
run = client.run_pipeline(experiment.id, "tf-flower", "tf-flower.tar.gz")

Custom model example

In [ ]:

# kfserving_op = components.load_component_from_url('https://raw.githubusercontent.com/kubeflow/pipelines/master/components/kubeflow/kfserving/component.yaml')
kserve_op = components.load_component_from_url(
    "https://raw.githubusercontent.com/kubeflow/pipelines/master/components/kserve/component.yaml"
)


@dsl.pipeline(name="KServe pipeline", description="A pipeline for KServe.")
def kservePipeline(
    action="apply",
    model_name="max-image-segmenter",
    namespace="anonymous",
    custom_model_spec='{"name": "image-segmenter", "image": "codait/max-image-segmenter:latest", "port": "5000"}',
):

    kserve = kserve_op(
        action=action,
        model_name=model_name,
        namespace=namespace,
        custom_model_spec=custom_model_spec,
    ).set_image_pull_policy("Always")


# Compile pipeline
compiler.Compiler().compile(kservePipeline, "custom.tar.gz")

# Execute pipeline
run = client.run_pipeline(experiment.id, "custom-model", "custom.tar.gz")