Simple notebook pipeline

Welcome to your first steps with Kubeflow Pipelines (KFP). This notebook demos:

Defining a Kubeflow pipeline with the KFP SDK
Creating an experiment and submitting pipelines to the KFP run time environment using the KFP SDK

Reference documentation:

Prerequisites: Install or update the pipelines SDK

You may need to restart your notebook kernel after updating the KFP sdk.

This notebook is intended to be run from a Kubeflow notebook server. (From other environments, you would need to pass different arguments to the kfp.Client constructor.)



In [ ]:

    
# You may need to restart your notebook kernel after updating the kfp sdk
!python3 -m pip install kfp --upgrade --user

Setup



In [ ]:

    
EXPERIMENT_NAME = 'Simple notebook pipeline'        # Name of the experiment in the UI
BASE_IMAGE = 'tensorflow/tensorflow:2.0.0b0-py3'    # Base image used for components in the pipeline



In [ ]:

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

Create pipeline component

Create a python function



In [ ]:

    
@dsl.python_component(
    name='add_op',
    description='adds two numbers',
    base_image=BASE_IMAGE  # you can define the base image here, or when you build in the next step. 
)
def add(a: float, b: float) -> float:
    '''Calculates sum of two arguments'''
    print(a, '+', b, '=', a + b)
    return a + b

Build a pipeline component from the function



In [ ]:

    
# Convert the function to a pipeline operation.
add_op = components.func_to_container_op(
    add,
    base_image=BASE_IMAGE, 
)

Build a pipeline using the component



In [ ]:

    
@dsl.pipeline(
   name='Calculation pipeline',
   description='A toy pipeline that performs arithmetic calculations.'
)
def calc_pipeline(
   a: float =0,
   b: float =7
):
    #Passing pipeline parameter and a constant value as operation arguments
    add_task = add_op(a, 4) #Returns a dsl.ContainerOp class instance. 
    
    #You can create explicit dependency between the tasks using xyz_task.after(abc_task)
    add_2_task = add_op(a, b)
    
    add_3_task = add_op(add_task.output, add_2_task.output)

Compile and run the pipeline

Kubeflow Pipelines lets you group pipeline runs by Experiments. You can create a new experiment, or call kfp.Client().list_experiments() to see existing ones. If you don't specify the experiment name, the Default experiment will be used.

You can directly run a pipeline given its function definition:



In [ ]:

    
# Specify pipeline argument values
arguments = {'a': '7', 'b': '8'}
# Launch a pipeline run given the pipeline function definition
kfp.Client().create_run_from_pipeline_func(calc_pipeline, arguments=arguments, 
                                           experiment_name=EXPERIMENT_NAME)
# The generated links below lead to the Experiment page and the pipeline run details page, respectively

Alternately, you can separately compile the pipeline and then upload and run it as follows:



In [ ]:

    
# Compile the pipeline
pipeline_func = calc_pipeline
pipeline_filename = pipeline_func.__name__ + '.pipeline.zip'
compiler.Compiler().compile(pipeline_func, pipeline_filename)



In [ ]:

    
# Get or create an experiment
client = kfp.Client()
experiment = client.create_experiment(EXPERIMENT_NAME)

Submit the compiled pipeline for execution:



In [ ]:

    
# Specify pipeline argument values
arguments = {'a': '7', 'b': '8'}

# Submit a pipeline run
run_name = pipeline_func.__name__ + ' run'
run_result = client.run_pipeline(experiment.id, run_name, pipeline_filename, arguments)

# The generated link below leads to the pipeline run information page.

That's it!

You just created and deployed your first pipeline in Kubeflow! You can put more complex python code within the functions, and you can import any libraries that are included in the base image (you can use VersionedDependencies to import libraries not included in the base image).

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.